In many storage systems, it is desirable to provide efficient release of storage capacity, for example, as such storage capacity becomes available due to deletion of storage volumes, in order to better accommodate the input-output (IO) processing demands of system users. However, conventional approaches to storage capacity release can be problematic. For example, in content addressable storage systems, errors in reference counts can lead to so-called “leaks” in physical capacity, where a given block of physical capacity has a non-zero reference count but is not actually utilized by any logical block and should therefore be deleted so as to be released for use by other logical blocks. These blocks of physical capacity are in effect “orphaned” within the storage system, and such orphaned physical capacity creates inefficiencies that can significantly undermine system performance.